WO2019185273A1

WO2019185273A1 - Redundant communication infrastructure for an ipv6 routing network based on a special router

Info

Publication number: WO2019185273A1
Application number: PCT/EP2019/054894
Authority: WO
Inventors: Andreas Foglar
Original assignee: Innoroute Gmbh
Priority date: 2018-03-24
Filing date: 2019-02-27
Publication date: 2019-10-03
Also published as: DE112019001534B4; DE112019001534T5; DE102018002461A1

Abstract

A network node with routing function for forwarding IPv6 packets based on their destination and source address is disclosed, which uses the IPv6 prefix having a particular structure for its routing decisions. The IPv6 prefix starts with a special indicator followed by 4-bit routing fields, which correspond to hierarchy levels of the network and. The respective fields in the destination address of incoming packets is used by the network nodes of the respective hierarchy level to forward the incoming packets to the corresponding Downlinks.

Description

Patent Application

Title

Redundant Communication Infrastructure for an IPv6 Routing Network based on a special Router.

Introduction

In the patent application DE102014000554 [1] an IPv6 Routing Network is described, which is based on a special Router. For redundancy issues so-called Crosslinks are described therein. Also, the configuration of the special router was done manually.

According to the present invention an IPv6 Routing Network is described, which can be deployed with multiple redundancy. It is based on a special Router with more than one Uplink, but without Crosslinks. The configuration occurs automatically, and optionally also manually. The correct interconnection is supervised by control protocols built into the Router. Instead of routing protocols the Router learns the optimum routes. In addition, the source address of the packets is checked by the Router.

MOTIVATION

Internet Router are communication nodes, which forward packets according to the Internet Protocol (IP) address. Today’s Routers have high complexity as well in the control plane as in the forwarding plane. In the control plane, complex Routing Protocols such as OSPF, RIP or BGP are required to determine the best Routes for every destination address, in the forward- ing plane, the fragmented address space of the version 4 of the Internet Protocol (IP4) requires routing tables with up to 500.000 entries and highly complex search functions for the Fong- est-Prefix-Match, which are most commonly realized by CAM (Content Addressable

Memory) integrated devices [2] This implies high current consumption and higher latency for the data transmission. The complex routing protocols of the control plane constitute a weak point for security, as potentially every data stream could be mis-routed or copied by manipu- lations from remote.

Introduction of Version 6 of the Internet Protocol (IP6) should at least solve the problem of the fragmented IP4 address space. IP6 defines a l28-bit wide address, for which among others various hierarchical or geographic address schemes have been proposed. In practice, none of them have been realized up to now. The reason for that is among others, that IP6 Router have been implemented consistently as extensions to IP4 Routers, which do not profit from hierar- chical address assignment. In the control plane additional protocols are added for IPv6, so that overall complexity is increased.

DISCLOSURE OF THE INVENTION

The idea of the new communication infrastructure is that each router evaluates only a limited part (e.g. 4-bit, 12-bit) of the routing address and forwards the IPv6 packets to a related num ber of output ports (e.g. 10, 16, 4096, 10.000). This requires a hierarchically structured IPv6 address.

The invention relies on the mapping of the international telephone numbers into the 64-bit prefix field of the IPv6 address. This option is already described in the patent US6868416 [3] of South Korea Telecom Authority. Here an address structure is chosen where the prefix field starting with the most significant bits - which are transmitted first over the transmission line - first contains a special indicator K with fixed bitmap (e.g. 4 or 8 bit) followed by e.g. 15 or 14 fields of 4 bit, respectively, where the telephone number is mapped to (Fig. 8).

In a preferred embodiment a special indicator with only 4 bits is chosen, as it leaves enough digits for the maximum 15 digits of an international telephone number according to the stand ard E.164 of the ITU-T [4] Hence every fixed telephone subscriber and also every mobile phone can be mapped. In the preferred embodiment the special indicator has the value 1010 (binary) or A (hexadecimal). Usage of a non-numeral character significantly differentiates the prefix address from the following digits. Moreover, this IPv6 address range is currently un used and far away from the multicast addresses which start with 1111 (binary).

An example for the mapping of a telephone number to the IPv6 prefix is shown in Table 1 : generation of the prefix of a company access in the regional network of Munich, Germany. The phone number of the company is +49-89-45241990. The last three of a total of 16 digits are available for private addressing within the company. They are marked in Table 1 with ‘xyz’. Thus, the company can address internally up to 1000 internal routers. Also, other map pings are thinkable, which are more clearly arranged at the expense of losing address range; for example: digit x could indicate the building, digit y the floor and digit z the room number where the sensor node is located. The sensors connected there are differentiated with the inter face identifier, the second 64-bit part of the IPv6 address.

able 1 : Prefix example Routers for this hierarchically structured address are significantly easier to implement that state-of-art routers. According to the present invention the new router has at least one Uplink port to higher level hierarchy and several Downlink ports to lower level hierarchy (Fig. 1).

The number of downlink ports determines how many 4-bit fields the router evaluates.

In the configuration process the router receives its own address, which starts with the special indicator followed by valid digits, their number depending on the respective hierarchy level. The remaining digits have the value zero.

For example, the top-level router for Germany has the address <A49>. For router addresses this short form is preferred. Alternatively the complete prefix can be given as it is also shown at the router display: <A49x.— .— .— >. It contains the router address A49, the digit x indi- cates the routing field and the minus signs indicate the non-relevant fields for this router (cf. Fig. 8). The router of the Munich company mentioned above would display

<A498.9452.4199. Ox— >. It evaluates one digit with the routing field‘x’, hence has 10 down link ports. Within the company for example a sensor node could be in the network where the address includes the complete 64-bit prefix (cf. Fig.9). The differentiation of terminals/sensors occurs according to the 64-bit interface identifier. A further example is a node (router) within the (public) Munich access network, which is capable to address up to 10.000 subscribers and accordingly evaluated four digits: its display would show <A498.9452.4xxx.x— >. Yet, a dis tributed access network with up to 100.000 subscribers could be considered as‘node’ in the sense of the present invention, as long as it behaves as a single node with the respective num ber of downlinks.

Datapath Functions of the Router

1. For packets, which are received at Downlink ports, the source address is checked in a first step. In case it does not match with the Router address, or the Routing field does not match with the number of the Downlink port from which the packet was received, the packet is discarded.

2. The address prefix of the destination address is compared to the router address.

3. In case of match, the routing field determines the Downlink port to which the packet is forwarded. If for example a Router has 16 Downlink ports a 4-bit field is evaluated.

4. In case the prefix of the packet address does not match with the Router address the for warding depends on the port where the packet has been received:

a. Packets from Downlink ports are forwarded to one of the Uplink ports.

b. The selection of the Uplink port is described below. c. Packets from Uplink ports are discarded (error case).

The discard measures described in point 1 increase the security, as the disables faking of false source addresses. For example, the Munich access network router mentioned above would check data packets from the company mentioned above if the prefix starts with A498.9452.4 and if the access port has the number 1990 (cf. Fig. 9). The check takes place accordingly in every router in upstream direction, when a data packet has been received at a Downlink port. If a data packet of the Munich company forwarded upstream - as the destination address is not in Bavaria - the German top-level router <A49x.— .— .— > would receive the packet at Downlink port 8 and check the destination address for the value <A498.x— .— .— >. The digits at the paces of the minus signs are not checked, these are transparent for the router. Checking parts of the source address at every Downlink port increases the security, as an attacker even when intruded within the network cannot fake arbitrary addresses.

The points 2., 3., 4. a. und 4.c. describe packet forwarding according to the destination address and are already claims of application [1] The case described under 4.c. normally should not occur. It is obviously a mis-routed data packet which is discarded. Point 4.b. includes the redundancy concept which is content of the present invention and which is described below.

The special Router according to the previous application [1] has only one single Uplink. It can be used for example to build the non-redundant network as shown in Fig. 2. By adding a further Uplink for example, the redundant network shown in Fig. 3 can be built. It is not redundant in the lowest hierarchy level: in case of a failure of a router in this hierarchy level all connected subscribers would be affected. In the second hierarchy level the network is twofold redundant, in the third hierarchy level it offers already four alternative routes. With every higher hierarchy level, the number of possible routes is doubled. Would the whole network with 15 hierarchy levels be built with Routers with two Uplinks and ten Downlinks it would reach the theoretical maximum of 2¹¹⁵ =16.384 top-level Routers. The number seems high, but in practice will be less.

As with the example of the Munich company above (cf. Table 1) the lower hierarchy levels are frequently within the private network of the subscriber. In the public part of the network usually is the access network, for example, the access node described above with four hierarchy levels, which is not redundant. Then, only the core network above the access network is implemented with redundancy, in this example starting with hierarchy level eight, and then the maximum number of top-level Routers reduces to 2⁸ = 256. These could be distributed worldwide, so that high redundancy is reached.

In case that the number is not sufficient some hierarchy levels could be occupied with Routers with three or four uplinks.

In particular, during the buildup phase of the network there will be regions which are not fully wired. This could imply for an individual Router that not every Uplink can reach all parts of the network. Hence the Router must take a decision over which Uplink a packet with a certain destination address must be forwarded. In the following a solution to that problem is described which has no need for a superordinate configuration instance or for routing protocols between nodes.

Routing in Upstream Direction

Router in redundant network must take a decision in Upstream direction, over which Uplink a packet shall be forwarded. This is the decision 4.b. mentioned above.

The following examples use Router with two Uplinks; the derived rules, however, are valid also for three or four uplinks.

In a fully equipped network as shown in Fig. 3 paths over all Uplinks reach the destination. In this case every Router distributes the packets cyclically over all available Uplinks. This distributes the load over the network.

In practice in particular in the initial phase the network will not be completely equipped. As example therefore a network is shown in Fig. 4 which results from the network of Fig. 3 when two routers and some connecting links are omitted. The position of the omitted routers and links is shown in Fig. 4 with dashed lines. From this figure these decision rules can be derived:

1. For the Routers <l-l>, <l-2>, <l-3>, <2-b> and <3b> the decision is easy, as only one Uplink is in operation.

2. The Routers <2-l>, <2-2> and <2-3> can reach the terminals with the addresses <l-x-x> only via Uplink A, the terminals with the addresses <3-x-x> only via Uplink B. The knowledge about the wiring in higher hierarchical layers is a priori not available within these Routers.

3. The Routers <3-l>, <3-2> and <3-3> can reach terminals with the addresses <2-x-x> only via Uplink B, terminals with the addresses <l-x-x> not at all. According to the present invention the individual Router learns this information from the source addresses of the packets received at the Uplink ports A and B. Therefor each Router creates a table which indicates the reachable address ranges for each Uplink. This prefix table is initialized with empty entries. When a packet is received at an Uplink port the router records the address range of the sender in the table. In case of the Routers <2-l>, <2-2> and <2-3> from Fig. 4 the prefix table will be populated as shown in Table 2, whereby the minus sign indicates an empty entry:

Table 2: Prefix table of the Routers <2-l>, <2-2> and <2-3> The Routers of the Group <3-l>, <3-2> and <3-3> contain Table 3:

Table 3: Prefix table of the Routers <3-l>, <3-2> and <3-3>

By inspection of these Router tables one can see if one part of the network can reach other address ranges not at all, non-redundant or redundant. If applicable, the network must be enhanced accordingly. This function is helpful for network administrators to find out the actual extension of the network. In addition, the Router can make available the prefix table accessible for remote readout via management protocols such as SNMP.

In a worldwide network each Router should at least learn which of the nine global zones [5] are reachable via which Uplink. An example how the prefix table of a Router could look like is shown in Table 4. Obviously, this Router is localized in Europe region, as zones 3 and 4 are reachable over both Uplinks, the American continent only over Uplink A and the rest of the world only over Uplink B.

Table 4: Example Prefix Table for Global Zones

The prefix table for Global Zones would be sufficient if each top-level Router within a Global Zone can reach each terminal within this Global Zone. Most likely this will not be the case. With high probability, however, the assumption will be true that each top-level Router within a country can reach each terminal within this country. If this statement is defined as a rule, then a prefix table with country prefixes in every Router would be always sufficient.

Currently about 300 country codes are defined [4] The memory needed to store this information is very small and can be realized with minimum access times.

With other definitions the prefix table would have to be adapted accordingly.

When learning prefixes, the fact is useful that the source addresses of the packets are verified.

Should a packet stream be forwarded in Upstream direction and all Uplink entries in the prefix table for a destination are empty the Router distributes the packets alternating over all Uplinks. As the same happens in return direction, after some time all reachable uplinks will be known.

It may happen that a Router falls out of service at a remote part of the network, so that this event is not notified locally. The respective entries in the prefix table would be outdated. As the algorithm distributes the packets to all Uplinks for an indefinite time a part of the packets would be lost. To avoid this case there are these options:

A. For all entries of the lookup table packets received from Uplinks are counted. If one of the counters stagnates while the other counter continues to count, obviously there must be a disruption. At a defined counter threshold, the entry for the stagnating Uplink will be set to‘not reachable’.

B. Within a larger timescale (e.g. one minute) alternatively one of the redundant prefix

entries will be cleared (set to‘not reachable’). In case that packets continue to be received from that Uplink the prefix entry will automatically be set again. This can be done in operation without interruption of the data packet stream. The more traffic prevails in that moment the faster the entry will be updated. In case of no traffic in that moment the Router could send test packets, preferably using existing protocols such as ICMR

Below the functions of the control plane are described.

While application [1] exclusively describes manual configuration of the Router via keys at the operation panel, the present invention defines a combination of configuration messages and manual configuration.

Basic ideas are to improve operation comfort and avoid configuration errors. When an Uplink port of a Router is connected with the Downlink port of another Router automatically the Router of the lower hierarchy level gets assigned the correct address from the higher hierarchy level.

For example, the top-level Germany Router with the address <A49> delegates the address <A498> to the Router connected at Downlink port 8, which becomes the top-level Bavaria district Router.

In the preferred embodiment this is done using available protocols such as DHCPv6 with the option Prefix Delegation [7] This is important in particular for the lowest hierarchy Routers where the terminals are connected.

In the preferred embodiment the special Router is equipped with keys and a display of 16 so- called 7-segment indicators to show address, routing field and don’t care digits (cf. Fig. 8). These indicators with their limited character set are used by intention to increase the impression of simplicity to the user. Every digit has assigned a key which by repeated pressing rotates through all characters allowed for that digit. Keys assigned to delegated address digits are automatically locked, for example the above-mentioned top-level Bavaria Router would lock the digits <A498>. With the other remaining keys, a user can select the desired final address, or just take over the delegated address. If for example initially for the whole Bavaria only subscribers of the Munich quarter‘Maxvorstadt’ are connected, then the address of the <A498> Router could be manually changed (downgraded) to <A498.945x.x— .— >, whereby after keying the two‘x’ characters automatically the remaining digits are set to (don’t care). The two‘x’ indicate that this Router has 100 Downlinks (cf. Fig. 8).

This rule also holds for Router where the Uplinks remain unconnected. The Router assumes to be a global top-level Router and automatically configures to <Ax— .— .— .— >. Except the key below the‘A’ all other keys are available to configure any address. The skipping of hierarchy levels can be useful, in particular in the setup phase of the network, to minimize installations.

In the preferred embodiment the Router does not accept misconfiguration. For example, if two downlink ports are connected (on the same or different Routers) an alarm is signaled, preferable by fast blinking of the affected ports at both sides, optionally supported by acoustic warning. Also, in such case the data transport over the involved ports is disabled.

A further misconfiguration is the wiring of the Router <3-l> shown in Fig. 5 with‘Error!’. Uplink A would configure the Router address <2-3>, while Uplink B has already configured address <3-l>. In this case Uplink port A signals the error (preferably optically and acoustically) and blocks data transport. In case port A was connected first Uplink B would signal the error.

An allowed configuration is shown in Fig. 5 at Router <l-2>: Uplink A configures address <l-2>, while Uplink B intends to configure address <l>. In such case, when two different addresses in the same hierarchy-branch are configured the Router always selects the lower hierarchy level, in this case <l-2>. Another example for hierarchy skipping is shown at Downlink 1 of Router <la>.

FIGURES

Fig. 1 shows the logical symbols for the Router with (exemplary) 1 and 2 Uplinks, respectively, and 3 Downlinks each. Aside, the arrows define the data flow directions upstream (from Downlink port to Uplink port) and Downstream (from Uplink port to

Downlink port). Within the logical symbol the Router address in brackets <adr>.

Fig. 2 shows a non-redundant network with 3 hierarchy levels, built with Routers according to Fig. 1 (left) with 1 Uplink and 3 Downlinks.

Fig. 3 shows a redundant network with 3 hierarchy levels, built with Routers from Fig. 1 (right) with 2 Uplinks and 3 Downlinks. The lowest Router hierarchy is non-redundant, the hierarchy level above is twofold redundant, the highest hierarchy level, consisting of the Routers <a>, <b>, <c> and <d> is fourfold redundant. The data transmission would still work even if three of these Routers fail.

Fig. 4 shows as example the partially populated network from Fig. 3, whereby the Routers <c> and <lb> are missing and accordingly their port wiring is missing, and in addition the links from Router <a> to Router <3a> and from <b> to <3a> are missing. In this example a part of the network is redundant, while other parts have partially redundant area parts, non- redundant area parts and not connected areas. The terminals of the group <l-x-x> have co connection with the group <3-x-x>.

Fig. 5 shows the network from Fig. 4 with two shortcuts, marked with ,Ok‘, which both skip a hierarchy level. In addition, an example false wiring is shown, marked with , Error!‘, where the original link from <3a> to <3-l> has been erroneously replaced by the link from <2b> to <3-l>. The Uplink A would delegate the address <2-3> which conflicts with <3-l>.

Fig. 6 shows the network from Fig. 2, where six Terminals are missing. Accordingly, the Routers <2-l> and <2-3> may be dropped, and also Router <2>. This example shows, that Router can be dropped, if only one of the Downlinks is used. Also, the hierarchy level can be skipped, but the Router below must be manually down-configured from the delegated address <2> to <2-2>. This is indicated with the hand symbol.

Fig. 7 shows the Router functional blocks of the data path (Uplink ports, Downlink ports, Switching Unit and Internal Packet Interface), the functional blocks for configuration

(Configuration Unit, Panel) and Memory Blocks (Prefix Table, Own address, Prefix Table).

Fig. 8 shows the partition of the IPv6 Address in Prefix und Interface Identifier of 64 Bit each, and how the Prefix in the preferred embodiment is divided into a Special indicator K and 15 fields of 4 bit each. Below two examples for Router configuration are shown with different size of Router address and Routing field. In example 1 Router address and Routing field together are smaller than 64 bits, and a don’t care field remains. Example 2 shows a special case that the Router address occupies the whole 64-bit prefix. This is the case for Routers in the lowest hierarchy level, where the terminals are connected. In such case the 64-bit Interface Identifier is used for routing. In the preferred embodiment the display shows the 64-bit prefix only.

Fig. 9 shows an example for a real, non-redundant communication infrastructure, starting with a global top-level Router with der address <Ax>, which for example could be localized at the UN-Organization ITU in Geneva. In Fig. 9 the short notation is used. The Router <Ax> has no Uplinks. At Downlink 4 the top-level Router of the global zone 4 is connected - located for example at the same place. Its Downlink 9 connected via a transmission link (e.g. SDH container, MPLS-pipe or a dedicated optical fiber) with the Uplink of the German top-level Router <A49>, which for example is located in Leipzig. Its Downlink 8 is connected via a transmission link with a Router in Munich, Maxvorstadt quarter, which has 100 Downlinks. As several Hierarchy levels in between are missing this Router is configured manually to address <A498.945xx>. Its Downlink number 24 is connected to the access node

<A498.9452.4xxx.x>, which has 10.000 Downlinks, usually implemented as ADSL or VDSL links. At Downlink port 1990 of the access node the company InnoRoute is connected, its Office-Router having the address <A498.9452.4199.0x> assigned. At Downlink 2 of the Office-Router a Sensor node with the address <A498.9452.4199.02l7> is connected, which e.g. is located in I^st floor, room 7. The other Downlinks of the Office-Router are available e.g. for PC networks. Also, the address of the Sensor node was manually corrected. The sensors connected to the Sensor node get assigned Interface Identifiers by address auto configuration according to DHCPv6, located in the second 64-Bit Feld of the IPv6 address.

Claims

1. A network node with routing function, in particular a Router and/or

an access node, and/or

a whole access network with routing function, and/or

an Office-Router, and/or

a sensor node,

for forwarding IPv6 packets based on their destination and source address, including: a number of bi-directional interfaces to network nodes of higher hierarchy level, also referred to as“Uplinks” in the following,

another, independent number of bi-directional interfaces to network nodes of lower hierarchy, also referred to as“Downlinks” in the following,

Memory to store the own routing address and the routing field position,

Prefix table to store address prefixes per Uplink,

Switching unit for data packets with access to own address memory, routing field memory and prefix table,

Panel for manual input and visual and acoustical output,

Internal packet interface to insert and extract control packets to and from the bi- directional data interfaces,

Configuration unit for own address and routing field with access to panel and internal packet interface,

wherein the router address always starts with the special indicator K, followed by a number of 4-bit fields depending on the hierarchy level,

wherein the routing field specifies the part of the destination address of the packet which is evaluated by the network node for the forwarding over the Downlinks and where the size is large enough so that all Downlinks can be addressed,

wherein the switching unit is enabled to forward a received data packet to an output depending on its own address, its internal prefix table and the routing field of the packet to a defined output,

wherein the internal packet interface is capable to receive control packets from all bi- directional data interfaces and transmit control packets to all bi-directional data interfaces.

2. A network node according to Claim 1, which can receive packets from all bi-directional data interfaces and which discards packets received at Downlinks and where the source address does not match the own address,

discards packets received at Downlinks, where the routing field relevant for that node does not match with the number of the received Downlink,

discards packets from Uplinks where destination address and router own address do not match,

forwards not discarded packets in case of match of destination address and router own address to a Downlink with the number taken from the packet’s routing field and forwards all other packets to one of the Uplinks,

wherein all comparisons are made between relevant own address fields of the router and the corresponding source or destination fields of the packets.

3. A network node according to the previous claims which forwards IPv6 packets where the 64-bit prefixes of source and destination addresses are structured with a special indicator K, followed by a number of 4-bit fields.

4. A network node according to claim 1 which maintains an internal prefix table,

which contains for each Uplink relevant parts of the source addresses of the packets which have been received for that Uplink

and which uses suitable methods to keep the entries up to date.

5. A network node according to claim 3 where the prefix table contains at least the global zones 1...9 of the country calling codes.

6. A network node according to claim 3, which forwards accepted packets to the uplinks wherein first, according to the prefix table these Uplinks are selected which contain relevant parts of the packet destination address

and then packets are distributed cyclically to these Uplinks,

wherein in case that the prefix table does not contain relevant parts of the destination address for any Uplink, packets are distributed cyclically to all Uplinks.

7. A network node according to claim 3 which continuously checks the actuality of all entries in the prefix table,

either by counting received packet per entry and per Uplink and in case of stagnation of one counter clearing the respective entry in the prefix table, or by cyclic tentative clearing of valid entries in large time scale, whereby in case of connectivity the respective entry will be set again in short time, but in case of non connectivity the respective entry will remain cleared.

8. A network node according to claim 1, which receives its address preferably via

configuration messages from other network nodes with higher hierarchy levels via the Uplinks, extends this address with 4-bit digit identifiers numbered per Downlink port and delegates the respective extended address to all Downlinks,

wherein the network node after power-up waits for prefix delegation messages from Uplinks and then configures its own address accordingly,

wherein in case of delegation of different addresses of the same hierarchical branch uses the lower hierarchy address for its own address and

in case of delegation of contradicting addresses of different hierarchy branches takes the first delegated prefix and ignores contradicting prefixes and blocks those Uplink ports for data packets from which the contradicting prefix delegations have been received,

wherein an optional extension of the delegated address can be done manually via the panel, wherein the delegated address range is blocked, but the address length can be extended, thereby moving the network node to a lower hierarchy level,

or in case of missing prefix delegation messages the network node address can be completely configured manually via the panel.

9. A network node according to claim 8 using standardized DHCPv6 packets for address configuration.